Wavelet Packet Analysis for Reducing the Effect of Spectral Masking to

Improve Auditory Perception

P. A. Dhulekar#1
#

Department of E & TC
Dr. B. A. Tech. University
Lonere 402103, Maharashtra, India
1
pravindhulekar@gmail.com

S.L.Nalbalwar#2

J.J.Chopade*3

Department of E & TC
Dr. B. A. Tech. University
Lonere 402103, Maharashtra, India
2
slnalbalwar@dbatu.ac.in

Department of E & TC
SNJBs LSKBJ COE
Chandwad, Maharashtra, India
3
jchopade@yahoo.com

membrane, which contains sensory hair cells. There are

about 3,000 to 3,500 inner hair cells, each with about 40
hairs and 20,000 to 25,000 outer hair cells, each with about
140 hairs. The tectorial membrane is situated above the
hairs.
The information about sounds is conveyed via inner hair
cells as the inner cells stimulate the afferent neurons. Outer
hair cells improve the basilar membrane responses thus
produce high sensitivity, and sharp the tuning (frequency
selectivity) to the basilar membrane [2].

Abstract Sensorineural hearing impaired person experiences

degraded frequency selectivity of the ear, due to increase in
spectral masking along the cochlear partition, results in
degraded speech perception. Splitting the speech signal by
means of filtering and down sampling at each decomposition
level, using wavelet packets with different wavelet functions,
helps to reduce the effect of spectral masking, through which it
improve speech intelligibility. This experiment was carried out
with vowel-consonant-vowel syllables for fifteen English
consonants; they are decomposed in eight frequency bands
with desired band selection. Dichotic presentation of the
processed speech signal showed improvement in speech quality,
recognition scores and less response time in the reception of
speech feature of voicing, place, and manner, indicating
usefulness of the scheme for better reception of spectral
characteristics.

Outer ear

Middle
ear

Inner ear

O ssicles

Vestibular apparatus
with semicircular
canals

Keywords- Sensorineural hearing impairment, Frequency

selectivity, Spectral masking, Wavelet packet, Dichotic
Presentation.

I.

Malleus
Incus
Stapes

Vestibular
nerve

Figure 1. Structure of the auditory system. Adapted from

Cochlear
Moore (1997), Fig.1-7
External

INTRODUCTION

Pinna

Cochlea or Inner ear is a snail shaped cavity like a spiral

shaped structure, filled with fluid. Fig. 1 shows the structure
of the auditory system. The start of the cochlea, where the
oval window is placed is known as the basal end and its
other end is known as the apical end [1]. The cochlea is
separated along its length into three chambers namely: scala
vestibuli, scala media, and scala tympani. Scala vestibuli
and scala media are split by a thin membrane called
Reissners membrane, and scala media is separated from
scala tympani by basilar membrane [1]. Incoming signal
causes the oval window for movement and pressure
differences at the tympanic membrane are applied, resulting
in cochlear fluid movement which provides upward and
downward movement of the basilar membrane. The
vibration occurs depends on the mechanical properties of
the basilar membrane and the frequency of the input signal
which fluctuate significantly from base to apex. The basilar
membrane is comparatively stiff and narrow at the basal end,
and less stiff and wider at the apex. The high frequency
signals produce maximum deflection at the base while low
frequency signals produce maximum deflection at the apex
[2]. Fig. 2 shows the vibration of the basilar membrane for
high, medium and low frequencies. The organ of Corti is
placed between the basilar membrane and tectorial

nerve

canal

Cochlea

Eardrum
(Tympanic
membrane)
Oval window
Round window
Eustachian tube

Nasal cavity

Fig. 1 Structure of the auditory system. Adapted from Moore (1997),

Fig.1-7
Frequency
4000

2000

1000

600

400

200

Rel. Amplitude

8000

10

15
20
Distance from stapes (mm)

25

30

35

Fig. 2 Amplitude patterns of vibration of basilar membrane for different

frequencies.

The impairment due to defect in the cochlea is known as

cochlear (sensory) impairment. Sensorineural hearing
impairment is caused by exposure to intense sound,

___________________________________
978-1-4244 -8679-3/11/$26.00 2011 IEEE

409

reduction in the effect of masking, so VCV syllables are

used.
For the evaluation of the speech processing strategies, a
set of fifteen nonsense syllables in VCV context with
consonants / p, b, t, d, k, g, m, n, s, z, f, v, r, l, y / and vowel
/a/ as in farmer were used. The features selected for study
were voicing (voiced: / b d g m n z v r l y / and unvoiced: / p
t k s f /), place (front: / p b m f v /, middle: / t d n s z r l /, and
back: / k g y /), manner (oral stop: / p b t d k g l y /, fricative:
/ s z f v r /, and nasals: / m n /), nasality (oral: / p b t d k g s z
f v r l y /, nasal: /m n /), frication (stop: / p b t d k g m n l y /,
fricative: / s z f v r /), and duration (short: / p b t d k g m n f v
l / and long: /s z r y /).

congenital defects leading to loss of cochlear hair cells,

damage to auditory neurons. The audiograms of the
sensorineural impairment show typical shapes depending on
the pathology such as high frequency hearing impairment,
elevated thresholds at low frequencies. There are different
impairments of moderate to severe bilateral sensorineural
hearing impairment, bilateral and profound deafness, and
unilateral very severe or profound deafness. Apart from the
elevated thresholds, sensorineural hearing impairment is
observed as by loudness recruitment, reduced frequency
selectivity and temporal resolution, and increased spectral
and temporal masking [3],[4].
Spectral masking or simultaneous masking is a
frequency-domain version of temporal masking, and tends
to occur in sounds with similar frequencies: a powerful
spike at 1 kHz will tend to mask out a lower-level tone at
1.1 kHz. It is masking between two concurrent sounds.
Sometimes called frequency masking since it is often
observed when the sounds share a frequency band e.g. two
sine tones at 440 and 450Hz can be perceived clearly when
separated. They cannot be perceived clearly when
simultaneous. In masking a sound is made inaudible by a
"masker", a noise or unwanted sound of the same duration
as the original sound [5],[6].
The discrete wavelet transform divide the signal spectrum
into frequency bands that are narrow in the lower
frequencies and wide in the higher frequencies [7], [8]. This
limits how wavelet coefficients in the upper half of the
signal spectrum are classified [9]. Wavelet packets divide
the signal spectrum into frequency bands that are evenly
spaced and have equal bandwidth and will be explored for
use in identifying transient and quasisteady-state speech.
WP are efficient tools for speech analysis, involve using
two-band splitting of the input signal by means of filtering
and downsampling at each decomposition level. Designing
the WP filter-bank involves choosing the decomposition tree
and then selecting the filters for each decomposition level of
the tree. For each decomposition level, there is a different
time-frequency resolution. Once the decomposition tree has
been selected, the next step involves selecting an
appropriate wavelet filter for each decomposition level of
the tree.
The decomposition of the input signal which includes
vowel-consonant-vowel for fifteen English consonants is
carried out by using Daubechies, Symlets, Biorthogonal
wavelets of different orders, a possible solution to problem
of spectral masking.
II.

B. The Speech Processing Strategies

For many signals, the low-frequency content is the most
important part. It is what gives the signal its identity. The
high-frequency content, on the other hand, imparts flavor or
nuance. Consider the human voice. If you remove the highfrequency components, the voice sounds different, but you
can still tell what's being said. However, if you remove
enough of the low-frequency components, you hear
gibberish.
In basic filtering process, the original signal passes
through two complementary filters and emerges as two
signals. Unfortunately, if we actually perform this operation
on a real digital signal, we wind up with twice as much data
as we started with. Suppose, for instance, the original signal
consists of 1000 samples of data. Then the resulting signals
will each have 1000 samples, for a total of 2000.These
signals A and D are interesting, but we get 2000 values
instead of the 1000.
C. Multiple Level Decomposition by DWT
There exists a more subtle way to perform the
decomposition using wavelets. By looking carefully at the
computation, we may keep only one point out of two in each
of the two 2000-length samples to get the complete
information. This is the notion of downsampling. We
produce two sequences called cA and cD, which includes
downsampling, produces DWT coefficients [8].
The decomposition process can be iterated, with
successive approximations being decomposed in turn, so
that one signal is broken down into many lower resolution
components. This is called the wavelet decomposition tree.
Fig. 3 shows multilevel decomposition by DWT up to level
3.

MATERIALS AND METHODS

A. The Speech Material

Earlier studies have used CV, VC, CVC, and VCV
syllables. It has been reported earlier that greater masking
takes place in intervocalic consonants due to the presence of
vowels on both sides [10], [11]. Since our primary objective
is to study improvement in consonantal identification due to

Fig. 3

410

Multilevel Decomposition by DWT up to Level 3

Wavelet packet method is a generalization of wavelet

decomposition that offers a richer range of possibilities for
signal analysis. In wavelet analysis, a signal is split into an
approximation and a detail. The approximation is then itself
split into a second-level approximation and detail, and the
process is repeated. For n-level decomposition, there are n+1
possible ways to decompose or encode the signal. This yields
more than 22^ (n-1) different ways to encode the signal [9].

D. Multiple Level Decomposition by Wavelet Packets

The problem with DWT is that it splits only
approximations at each decomposition level while in
wavelet packet analysis, the details as well as the
approximations can be split also approximation coefficient
and detail coefficient for both low and high frequency are
found out. Fig. 4 shows multilevel decomposition by
Wavelet Packets up to level 3.

III.

EXPERIMENTAL RESULTS

In this analysis we have decomposed fifteen nonsense

syllables in VCV context with consonants / p, b, t, d, k, g, m,
n, s, z, f, v, r, l, y / and vowel /a/ as in farmer, using wavelet
packets with different wavelets like Daubechies, Symlets and
Biorthogonal. Fig. 5 shows the Multilevel Decomposition of
VCV context /ada/ which is obtained by sym5 gives
Approximation & Detail Coefficients. Decomposition Tree
with data at node for VCV context /ada/ by sym5 is shown
in Fig. 6.The perfect reconstruction error is shown in Table I
TABLE I
Discrete Wavelet Transform
Fig. 4

Wavelet Packet

Multilevel Decomposition by Wavelet Packets up to Level 3

Fig. 5

6.6877e-011
9.7225e-012

Decomposition of context /ada/ up to level 3 with Approximation coefficients and Detail coefficients using sym5

411

Fig. 6

IV.

Decomposition tree with data for node of context /ada/ up to level 3 using sym5

For reducing the effect of increased spectral masking, speech

processing schemes based on spectral splitting of speech
signal, by using comb filters with complementary pass bands,
for binaural dichotic presentation have been reported earlier.
Splitting the speech signal and compressing the frequency
bands is a possible solution to reduce the effect of increased
masking. In this study, a processing scheme of speech signal,
by using discrete wavelet transform with different types of
wavelets with different orders has been investigated to
decompose the speech signal into two for binaural dichotic
presentation, presented to the two ears which may help in
improving the perception of various consonantal features.
The speech signals were decomposed with discrete wavelet
transform at various levels to get low frequency and high
frequency signal components. For each decomposition level,
there is a different time-frequency resolution. Once the
decomposition tree has been selected, the next step involves
selecting an appropriate wavelet type depending on
orthogonally, symmetry such as Daubechies, Symlets and
Biorthogonal wavelet functions. It helps in reducing
temporal masking and spectral masking simultaneously,
thereby improving the speech perception.

DISCUSSION

Sensorineural hearing impairment is associated with

increased masking. Increased spectral masking associated
with broad auditory filters results in smearing of spectral
peaks and valleys, and leads to difficulty in perception of
consonantal feature. Increased forward and backward
temporal masking of weak acoustic segments by strong ones
causes reduction in discrimination of voice-onset-time,
formant transition, and burst duration that are required for
consonant identification. Thus the overall effect of two
types of masking is a difficulty in discrimination of
consonants, resulting in relatively degraded speech
perception by persons with sensorineural impairment.
Masking takes place primarily at the peripheral auditory
system. In speech perception, the information received from
both the ears gets integrated. Hence splitting of speech
signal into two complementary signals such that signal
components likely to mask each other get presented to the
different ears can be used for reducing the effect of
increased masking. This technique can be used for
improving speech reception by persons with moderate
bilateral sensorineural impairment, i.e. residual hearing in
both the ears.
This study shows that, hearing impaired subjects are able
to perceptually integrate the dichotically presented speech
signal. Processing schemes for splitting the speech signal in
a complementary fashion to reduce the effects of increased
masking at the peripheral level improved speech reception.
The dichotic presentation also decreases the load on the
perception process. For hearing impaired subjects, the
improvement in consonantal reception and reduction in
response time do not follow the same trend.

REFERENCES
[1]
[2]
[3]
[4]

412

B. C. J. Moore, An Introduction to Psychology of Hearing. 4th ed.

London: Academic, 1997.
B. C. J. Moore, Speech processing for hearing impaired:
successes, failures, and implications for speech mechanisms,
Speech Communication., vol. 41, pp. 81-91, 2003
L. R. Rabiner and R. W. Schafer, Digital Processing of the Speech
Signals, Englewood Cliffs, NJ: Printice Hall, 1978.
T. Arai, K. Yasu, and N. Hodoshima, Effective speech processing
for various impaired listeners, Proc. 18th International Congress
Acoustics (ICA), 2004, pp. 1389 - 1392.

[5]

[6]

[7]

P.N.Kulkarni and P.C.Pandey,Optimizing the comb filters for

spectral splitting of
speech to reduce the effect of spectral
masking, in Proc. International conference on signal processing,
Communication and Networking (MIT,Chennai,India), pp.6973,Jan 4-6, 2008.
A. N. Cheeran, and P. C. Pandey, Evaluation of speech processing
schemes using binaural dichotic presentation to reduce the effect of
masking in hearing-impaired listeners, in Proc. 18th International
Congress on Acoustics (ICA 2004, Kyoto, Japan), pp. 1523 - 1526,
Apr. 49, 2004.
I. Cheikhrouhou, R.B. Atitallah, K. Ouni, A.B. Hamida, N.
Mamoudi,and N. Ellouze, Speech analysis using wavelet
transforms dedicated to cochlear prosthesis stimulation strategy,
1st Intern. Symp. On Control, Communications and Signal
Processing, 2004, pp. 639642.

[8]

[9]
[10]

[11]

413

Alessia Paglialonga, Gabriella Tognola, Giuseppe Baselli, Marta

Parazzini, Paolo Ravazzani, and Ferdinando Grandori Speech
Processing for Cochlear Implants with the Discrete Wavelet
Transform: Feasibility Study and Performance Evaluation Proc.
28th IEEEEMBS Annual International Conference, New York City,
USA, Aug 30-Sept 3, 2006 pp.3763-3766.
S. Mallat, A theory for multiresolution signal decomposition: the
wavelet representation, IEEE Trans. on Patt. Anal. Machine Intell,
vol. 11(2), pp. 674-694, 1989.
Chaudhari D. S. and Pandey P. C. Dichotic presentation of speech
signa1 using critical filter bank for bilateral sensorineural hearing
impaired. Proceedings of I6* International Congress on
Acoustics, Seattle, Washington, 1998, vol. 1, pages 213-214
D. S. Chaudhari, and P. C. Pandey, Dichotic presentation of
speech signal with critical band filtering for improving speech
perception, in Proc. ICASSP 98, Seattle, Washington,
USA,pp.3601-3604.